Hindawi Publishing Corporation Computational and Mathematical Methods in Medicine Volume 2014, Article ID 813067, 19 pages http://dx.doi.org/10.1155/2014/813067

Research Article A Supervised Network Analysis on Expression Profiles of Breast Tumors Predicts a 41-Gene Prognostic Signature of the Transcription Factor MYB across Molecular Subtypes

Li-Yu D. Liu,1 Li-Yun Chang,2 Wen-Hung Kuo,3 Hsiao-Lin Hwa,2 King-Jen Chang,3,4 and Fon-Jou Hsieh2,5

1 Biometry Division, Department of Agronomy, National Taiwan University, Taipei 106, Taiwan 2 Department of Obstetrics and Gynecology, College of Medicine, National Taiwan University, Taipei 100, Taiwan 3 Department of Surgery, College of Medicine, National Taiwan University, Taipei 100, Taiwan 4 Cheng Ching General Hospital, Taichung 400, Taiwan 5 Research Center for Developmental Biology and Regenerative Medicine, National Taiwan University, Taipei 100, Taiwan

Correspondence should be addressed to Fon-Jou Hsieh; [email protected]

Received 29 May 2013; Revised 7 October 2013; Accepted 20 October 2013; Published 3 February 2014

Academic Editor: Seiya Imoto

Copyright © 2014 Li-Yu D. Liu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background. MYB is predicted to be a favorable prognostic predictor in a breast cancer population. We proposed to find the inferred mechanism(s) relevant to the prognostic features of MYB via a supervised network analysis. Methods. Both coefficient of intrinsic dependence (CID) and Galton Pierson’s correlation coefficient (GPCC) were combined and designated as CIDUGPCC. It is forthe univariate network analysis. Multivariate CID is for the multivariate network analysis. Other analyses using bioinformatic tools and statistical methods are included. Results. ARNT2 is predicted to be the essential gene partner of MYB. We classified four prognostic relevant gene subpools in three breast cancer cohorts as feature types I–IV. Only the probes in feature type II are the potential prognostic feature of MYB. Moreover, we further validated 41 prognosis relevant probes to be the favorable prognostic signature. Surprisingly, two additional family members of MYB are elevated to promote poor prognosis when both levels of MYB and ARNT2 decline. Both MYBL1 and MYBL2 may partially decrease the tumor suppressive activities that are predicted to be up-regulated by MYB and ARNT2. Conclusions. The major prognostic feature of MYB is predicted to be determined by the MYB subnetwork (41 probes) that is relevant across subtypes.

1. Introduction Miao et al. described a transient defect in mammary gland development in the mouse model with the genetic deletion Breast cancer (BC) has become a global health problem of MYB.TheysuggestedthatMYB is critical for tumor among women in recent years. Finding more cost effective growth and mammary carcinogenesis [2]. MYB transcription strategies to better control this problem is highly desirable. factors (TFs) in the MYB family are widely distributed in It should be noted that the nature of heterogeneity for eukaryotic organisms [3, 4]. MYB family members consist of breast cancer at molecular and clinical level [1]stillremains A-, B-, and C-MYBsindiversevertebrates.A-MYB (MYBL1) challenging to breast cancer care and prevention. plays a critical role in mammary gland development. In We selected MYB for the global network study because female mouse model, MYBL1 is expressed in breast duc- it is essential for mammary gland development and tumori- tal epithelium, mainly during pregnancy-induced ductal genesis [2]. However, the most important reason was that branching and alveolar development [5]. B-MYB (MYBL2), a our preliminary data suggested MYB to be a good prognostic mitotic regulator, could be implicated in breast tumorigenesis predictor among 181 infiltrating ductal breast carcinomas because it is detected in a wide variety of cancer cells and plays basedonKaplan-Meiersurvivalanalysis. anessentialroleduringcellcycleprogression[6–8]. It has 2 Computational and Mathematical Methods in Medicine been documented that C-MYB (MYB) plays different roles in ER(−)PR(+)HER(−) (5A), ER(−)PR(+)HER(+) (6A), and normal and cancer cells [9]. The findings of Thorner et al.[10] ER(−)PR(+)HER(?) (3A). In addition, we designated the indicated that C-MYB may not be behaving as an oncogene cohort of Groups IE and IIE containing ninety gene expres- in estrogen receptor positive (ER(+)) luminal breast tumors sion profiles as 90A cohort. The cohort from subcohorts and suggested that it may be behaving as a tumor suppressor for TN, ERBB2+, ER(−) PR(+)HER(−), ER(−)PR(+)HER(+), in this disease. All these findings described above indicate and ER(−)PR(+)HER(?) to make 91 gene expression profiles the important roles of MYB family members in relation to was designated as 91A cohort. For 181A cohort, it includes mammary gland and breast cancer development in the model 90A cohort and 91A cohort. systems. However, more studies in a genome-wide scale for finding the roles of MYB in breast cancers are essential to fill 2.2. Microarray Data Analyses. Aglobalviewofagene in the gaps for the current findings in the field. profileperbreasttumorspecimenwasanalyzedusingHuman Thisstudyaimedatreassessingthedevelopmentally 1A (version 2) oligonucleotide microarray (half a genome important transcription factor MYB mediated transcriptional size: 22 k) (Agilent technologies, USA). The heatmaps were regulatory networks in relation to breast cancer development displayed after unsupervised hierarchical clustering. For and clinical outcome. unsupervised hierarchical clustering, the log2 ratio for each gene was first centered by subtracting the median across 2. Materials and Methods all samples to discriminate the subclass of the dataset. The “hcluster” function in “stats” package was utilized to perform 2.1. Features of Surgical Specimens for Generating the Dataset the unsupervised clustering. We used the Euclidean distance of Gene Expression Profiles. We used immunohistochemical and the complete linkage as the default settings. Then, the (IHC) statuses for three biomarkers (i.e., estrogen receptor 𝛼 selected gene expression profiles were fed into the software (ER), progesterone receptor A (PR), and HER-2/neu (HER)) R2.15.1 for displaying gene list (Y axis) that is derived from as the classifiers to identify eight intrinsic subtypes. However, hierarchical clustering analysis on the gene profiles of selected for ERBB2 (IHC score: 2+), determination of Her-2/neu gene arrays (X axis)togeneratetheheatmaps.Theheatmapwas copy number by chromogenic in situ hybridization (CISH) produced by “rect” function to make the customized view was performed [12]. As such, IHC/CISH status was used for of the subcohorts. In addition, we used Gene Spring GX7.3.1 determining HER status. (Agilent Technologies, USA) for generating Venn diagrams Ninety specimens of primary infiltrating ductal breast andforretrievingupdatedgeneannotation.ANOVAhasthe carcinomas (IDCs) consist of group IE (i.e., ER(+)PR(+)) advantage of performing both dichotomous and multichoto- (61/90) and group IIE (i.e., ER(+)PR(−)) (29/90). mous analyses. ANOVA test for the relationship between Ninety-one samples of IDCs consist of triple negatives mRNA levels of MYB and the statuses of a clinical index of (TN) (i.e., ER(−)PR(−)HER(−)) (48/91), ERBB2+ (i.e., interest in a given population as well as the statistical methods ER(−)PR(−)HER(+)) (29/91), ER(−)PR(+)HER(−)(5/91), for establishing MYB transcriptionalregulatorynetworkwere ER(−)PR(+)HER(+) (6/91), and ER(−)PR(+)HER(?) (3/91). described previously [11–14].Weusedthesamedataanalyses Those samples were obtained from patients who underwent described above for analyzing other transcription factors of surgeryatNationalTaiwanUniversityHospital(NTUH) interest. We performed Kaplan-Meier survival analyses [15] between 1995 and 2007. The tumor samples for this study using “survival” package in R (version 2.15.1) using the gene were the remaining frozen samples from diagnostic purpose. profiles of 90A cohort, 91A cohort, and 181A cohort or the All patients provided informed consent according to the extracted gene pools of interest in the assigned cohorts. guidelines approved by the Institutional Review Board To quantify the weight of hazard ratios associated with (IRB) at NTUH (IRB number: 200706039R, Research Ethics prognostic gene signature and the traditional prognostic Committee at National Taiwan University Hospital, Taipei, factorsinagivencohortofinterest,bothunivariateand Taiwan). The survival status of this cohort was derived from multivariate COX proportional hazard (COXPH) regression the recent medical recording collected in 2011 (by WHK). models in R package (version 2.15.1) were performed. Other medical records of the patients were obtained from the great assistance from the office of medical record (Cancer 2.3. Features of the Adapted Network Analysis Based on the Registry, Medical Information Management Office, NTUH). Dataset of 181 Gene Expression Profiles. The DNA microarray At the time of this study, the record of cancer treatments for becomes mainstay technique used in medical research. In these cancer patients was only partially complete. recent years, we have designed the network analysis for full The microarray data for this study (181 gene expression prediction of a transcription network for a given transcription profiles) have been submitted to the NCBI Gene Expression factor in a population of interest [11, 13]. However, we present Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) under the partial results predicted by a supervised network analysis accession number GSE24124. In this study, we designated mainly due to the existing limitations in this dataset of ours 90A as the gene expression microarray dataset for 90 ER(+) described in Sections 2.1 and 2.2.Asupervisednetwork breast tumors. It consists of two subsets that are group analysis approach was developed [12]. IE (61A) and group IIE (29A). The definition for 91A is We initially designed the network analysis approach the gene expression microarray dataset of 91 ER(−) breast including IHC stain to guide the network prediction, in part tumors. It consists of subsets for TN (48A), ERBB2+(29A), [13]. The predicted numbers of human putative transcription Computational and Mathematical Methods in Medicine 3 factors genome-wide are between 1,850 and 4,105 [16]. It is feature of MYB based on the gathered evidence from the impossible to provide IHC stain for each transcription factor inferred transcriptional regulatory network of MYB in rela- of interest using clinical tumor samples within the same tion to biochemical phenotypes, malignant phenotypes, and cohort. Therefore, we used the data at mRNA level to find the supporting evidence from others. Fourth, we further selected inferred target for a TF of interest as the rule of thumb. the overlapping gene set in the given feature type of both Statistically, to deal with continuous variables of interest, 90A cohort and 181A cohort to be the consensus prognostic CIDhastheadvantagetomeasurethesubCIDvalueof signature of MYB.Finally,theprognosticsignaturerelevantto asubgroupwithasmall𝑁 number (𝑛≒10)without clinicopathological parameter(s), subtype(s), and treatment increasing the statistical errors. Biologically, we chose the response(s) may be concluded from this study. 1/10th subgrouping strategy (𝑛≒10) among tested sub- grouping strategies although each transcription factor of 3. Results and Discussion interest may need adjust subgrouping strategy for network analysis to increase both sensitivity and specificity of network Our evaluation focused on the clinical pathophysiologi- analysis. However, we constantly compare subtype relevant cal and/or subtypical implication of MYB. Such genome- transcriptional regulatory events that are normally in a small wide transcriptional activities might offer us new insights 𝑛≒30 sample group ( )inourmodelsystem.Itisreasonable into the clinical behaviors of breast cancers, such as their to set 1/10th subgrouping as the best of choice. responsiveness to standard cancer therapies and survival after surgical removal of breast tumor(s). For instance, both network prediction and some validated evidence suggested a 2.4. Experimental Design. The 181 gene expression profiles transcription factor STAT3 to be a center regulator of estrogen of the human infiltrating ductal breast carcinoma contain receptor negative (ER(−)) breast cancers [12]. Many potential more than eight breast cancer intrinsic subtypes based on and existing drug targets or genes resistant to standard cancer IHC/CISH results. This offers the opportunity of finding treatments have been identified via an established scheme prognostic relevant gene pools among breast cancer subtypes. for the network analysis [12]. Such new approach allows In this model system, Kaplan-Meier survival analysis [15] more valuable information to be extracted and they can be predicts MYB to be a favorable prognostic predictor in linked together to form functional networks. These inferred 181A cohort. In this study, the rationale for selecting 90th transcriptional regulatory networks in a clinical breast cancer percentile as the cut-off point for Kaplan-Meier survival model system are expected to assist us in unraveling the iden- analysis is mainly to match the subgrouping strategy of both tity of breast cancer subtypes at molecular level. Meanwhile, univariate and multivariate CID. the options for both cancer prevention and cancer treatment The transcriptional regulatory network analysis is highly of different breast cancer subtypes may be indicated via this sensitive in measuring both the existing and novel gene study.Finally,thediscoveryofthegenesignature,whichis expression relationships between a TF and its potential prognostically relevant in a subset of highly MYB expressed target gene in a population of interest [13]. In addition, breasttumors,isexpectedtobeachieved. the gene expression relationships between the combinatorial interacted TFs (𝑁≧2) and their potential shared target gene in a given population are measured [11]. Here, we designed 3.1. The Most Relevant Transcriptional Regulatory Event of a combined strategy including both network analysis and MYBinRegulatingGenesforPredictingClinicalOutcome Kaplan-Meier survival analysis to find the prognostic relevant Is Neither Subtype Dependent Nor Unique Clinical Param- transcriptional regulatory subnetwork of MYB.Theprognos- eter Dependent tic values of these network components (i.e., probes) were further predicted by Kaplan-Meier survival analysis. Thus, 3.1.1. The Potential Clinical Impact of MYB as a Tumor such strategy allows 10% population to be selected by their Suppressor and Its Relation with Favorable Prognostic Features relevance to both the inferred transcriptional event and prog- of MYB. The mRNA levels of MYB are relatively high in nosis. For the transcription factor MYB,weproposedthatthe Group IE of ER(+) IDCs as compared to other subtypes in most relevant subnetwork of MYB with the highest subCID 181IDCs (Figures 1(c) and 1(d)). Typically, an increased gene value in a subset of tumor samples may be co-localized expression of MYB is relevant in PR(+) IDCs (see results of with the top 10% tumor sample population that expresses ANOVA tests in Figures 1(a) and 1(b)). We predicted that the high levels of MYB and indicates a favorable prognosis. clinicaloutcomeinthesesubtypesmaybefavorableduetothe Meanwhile, we added a few key steps to control confounders action of MYB, in part. In addition, MYB is one of the deter- andtoquicklylocatethemajorprognosticfeaturesofMYB. minants for early tumor development in clinicopathological First, we chose three populations of interest (i.e., 91A cohort, features of lymphovascular invasion (LVI), histological grade 90A cohort, and 181A cohort) and classified their prognostic (Grade or G), tubule formation (TF), nuclear pleomorphism predictors into four types based on their differential relevance (NP), tumor size (size), and the number of lymph node among these populations. Second, we identified the subpool metastasis (LNM) in 90A cohort (Figure 1(a)). It is also the of genes that is not only a subpool of the prognostic predictors significant determinant of early statuses for G, MC, NP, and of a given type but the components of MYB inferred network. TF in 181A cohort (Figure 1(b)). They were classified as genes with a given feature type. Third, To prove the clinical behavior of MYB due to its func- we proposed those genes to represent the major prognostic tion as the transcription factor, a series of analyses was 4 Computational and Mathematical Methods in Medicine

2 2 2 2 2 2 P = 0.041 P = 0.826 P = 0.05 P = 0.504 P = 0.028 P = 0.023

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 MYB (5586) expression −2 −2 −2

−3 (29) (61) −3 (42) (30) −3 (20) (56) −3 (25) (61) −3 (25)(29)(14)(17) −3 (26) (47) (13) − + − + − + − + 0123 1 23 & 4 PR HER LVI LYM LNM Tumor size

2 2 2 2 2 2 P = 0.055 P = 0.003 P = 0.049 P < 0.001 P = 0.463 P = 0.871 1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2 MYB (5586) expression

−3 (13) (38) (35) −3 (23) (43) (17) −3 (2) (29) (48) −3 (6) (46) (27) −3 (46) (22) (11) −3 (38) (52) 1 2 3 & 4 12 3 123 123 12 3 <48 >48 Stage Grade TF NP MC Age (a)

P < 0.001 P = 0.046 P = 0.284 P = 0.158 P = 0.37 P = 0.502 1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2

−3 −3 −3 −3 −3 −3 MYB (5586) expression −4 −4 −4 −4 −4 −4 (106) (75) (95) (65) (57) (101) (67) (107) (69)(51)(26)(28) (53) (95) (28) − + − + − + − + 0123 123 & 4 PR HER LVI LYM LNM Tumor size

P = 0.878 P < 0.001 P = 0.002 P < 0.001 P < 0.001 P = 0.081 1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2

−3 −3 −3 −3 −3 −3 MYB (5586) expression −4 −4 −4 −4 −4 −4 (35) (81) (61) (24) (77) (69) (3) (39) (120) (7) (75) (80) (68) (51) (43) (64) (117) 123 & 4 111123 23 23 23 <48 >48 Stage Grade TF NP MC Age (b)

Figure 1: Continued. Computational and Mathematical Methods in Medicine 5

P value < 0.001 2 P value < 0.001 2

0 0

−2 −2 MYB (5586) expression MYB (5586) expression

−4 −4

IEIIE ERBB2+ TN IE Non-IE

(c) (d) MYB (5586) mRNA expression 1 P = 0.026

0.8

0.6

0.4

Surviving fraction 0.2

0 0 5 10 15 20 Time to death (years) Low (162) High (19) (e)

Figure 1: Clinical impact of MYB in two cohorts of infiltrating ductal breast carcinomas. ANOVA test results of MYB (5586) mRNA levels in eight clinical indices, progesterone receptor (PR), HER-2/neu (HER), lymphovascular invasion (LVI), lymph nodal category (lymph node metastasis status (LYM) and number of lymph node metastasis (LNM)), age, tumor size (size), grade (nuclear pleomorphism, mitotic count, and tubule formation), and cancer stage in 91A cohort and 181A cohort, respectively ((a) and (b)). Lower panel, box plot analysis of MYB (5586) mRNA levels in four subtypes (i.e., Groups IE, IIE, triple negatives (TN), and ERBB2+) (c) and in two types (i.e. group IE (IE) and non-group IE(non-IE)) (d). The red dot line within box stands for the mean value for each subgroup in the plot. The black line withinbox stands for the median value for each subgroup in the plot. The survival analysis (Kaplan-Meier survival analysis) on a breast tumor group with high MYB mRNA levels versus the group with low MYB mRNA levels in 181A cohort (e). The Agilent feature number for MYB is 5586.

performed to dissect the major action of MYB via the MYB on breast cancers which is significant in both 90A cohort transcriptional regulatory network approach. As a result, the and 181A cohort (Figure 3(a)). The results from Venn diagram predicted tumor suppressive activities of MYB may be due to analysis further demonstrate the most relevant activities its transcriptional activities. These activities of potential MYB of MYB in coupling with ARNT2 via three networks of target genes are overlapping with some favorable prognostic MYB ARNT2 in 90A cohort and 181A cohort, respectively predictors in the tested cohorts. (Figure 3(b)). The gene profiling of clinical relevance and Two clinically relevant MYB clusters and the predicted of involved signal transduction pathways for two net- MYB transcriptional activities suggest ARNT2 to be the works at the lower panel is demonstrated by bar charts obligate gene partner of MYB (Figures 3(a) and 3(b)). It suggesting the tumor suppressive effect of MYB (Figures potentially co-contributes with MYB for its clinical impact S6.1–S6.9, in Supplementry Material available online at 6 Computational and Mathematical Methods in Medicine

3 P = 0.662 3 P = 0.4663 P = 0.0313 P = 0.9563 P = 0.0033 P = 0.021

2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1 ARNT2 (3187) expression

−2 −2 −2 −2 −2 −2 (29) (61) (42) (30) (20) (56) (25) (61) (25)(29)(14)(17) (26) (47) (13) − + − + − + − + 0123 1 23 & 4 PR HER LVI LYM LNM Tumor size

3 P = 0.0883 P < 0.0013 P = 0.0183 P < 0.0013 P = 0.5043 P = 0.219

2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1 ARNT2 (3187) expression −2 −2 −2 −2 −2 −2 (13) (38)(35) (23) (43) (17) (2) (29) (48) (6) (46) (27) (46) (22) (11) (38) (52) 123 & 4 123 123 123 123 <48 >48 Stage Grade TF NP MC Age (a) 3 3 3 3 3 3 P = 0.709 P = 0.42 P = 0.025 P = 0.845 P = 0.003 P = 0.018 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1 ARNT2 (3742) expression

−2 (29) (61) −2 (42) (30) −2 (20) (56) −2 (25) (61) −2 (25)(29)(14)(17) −2 (26) (47) (13) − + − + − + − + 0123 1 23 & 4 PR HER LVI LYM LNM Tumor size

3 3 3 3 3 3 P = 0.099P < 0.001 P = 0.026 P < 0.001 P = 0.413 P = 0.211 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1 ARNT2 (3742) expression

−2 (13) (38) (35) −2 (23) (43) (17) −2 (2) (29) (48) −2 (6) (46) (27) −2 (46) (22) (11) −2 (38) (52) 123 & 4 123 123 123 123 <48 >48 Stage Grade TF NP MC Age (b)

Figure 2: Continued. Computational and Mathematical Methods in Medicine 7

P < 0.001 P = 0.408P = 0.222 P = 0.788 P = 0.111 P = 0.042 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2 ARNT2 (3187) expression (106) (75) (95) (65) (57) (101) (67) (107) (69)(51)(26)(28) (53) (95) (28) − + − + − + − + 0123 1 23 & 4 PR HER LVI LYM LNM Tumor size

P = 0.342 P < 0.001P < 0.001 P < 0.001 P = 0.054 P = 0.465 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2 ARNT2 (3187) expression (35) (81) (61) (24) (77) (69) (3) (39)(120) (7) (75) (80) (68) (51) (43) (64) (117) 1 2 3 & 4 123 123 123 123 <48 >48 Stage Grade TF NP MC Age (c)

P < 0.001 P = 0.535 P = 0.283 P = 0.505 P = 0.115 P = 0.093 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2 ARNT2 (3742) expression (106) (75) (95) (65) (57) (101) (67) (107) (69)(51)(26) (28) (53) (95) (28) − + − + − + − + 0123 1 23 & 4 PR HER LVI LYM LNM Tumor size

P = 0.681 P < 0.001 P < 0.001 P < 0.001 P = 0.02 P = 0.476 2 2 2 2 2 2

1 1 1 1 1 1

0 0 0 0 0 0

−1 −1 −1 −1 −1 −1

−2 −2 −2 −2 −2 −2 ARNT2 (3742) expression (35) (81) (61) (24) (77) (69) (3) (39)(120) (7) (75) (80) (68) (51) (43) (64) (117) 1 2 3 & 4 123 123 123 123 <48 >48 Stage Grade TF NP MC Age (d)

Figure 2: Clinical impact of ARNT2 in two cohorts of infiltrating ductal breast carcinomas. ANOVA test results of ARNT2 (3187) and ARNT2 (3742) mRNA levels in eight clinical indices are shown in 90A cohort ((a) and (b)) and 181A cohort ((c) and (d)), respectively. There are two probes for ARNT2. They have the Agilent feature numbers 3187, 3742 and respectively. 8 Computational and Mathematical Methods in Medicine

MYB is a favorable prognostic factor in 181A cohort.

GTFNP181 MC 181

The clinically significant 390 1062 5409 probes in 181A cohort.

22 TFs or their subunits/variants are in a MYB cluster of 181A cohort MYB is an ER 𝛼 target gene and high MYB level is predominantly found in 90A cohort

The clinically significant probes in 90A cohort

3 common probes in a MYB cluster of 90A cohort are identified (i.e. MYB, ARNT2, and ARNT2 variant) (a) MYB clusters Cohort MYB ARNT2 CS MYB ARNT2 Cohort CS MYB ARNT2 Univariate Multivariate Univariate Multivariate Cohort CS

90 IDCS

4,982 probes 727 probes 480 probes

Univariate Multivariate Univariate Multivariate Cohort CS

181 IDCs

6,603 probes 2,922 probes 2,727 probes (b) MYB networks in two cohorts The relevant determinants of The relevant determinants of clinical The relevant signal transduction The relevant signal transduction clinical parameters in parameters in MYB ARNT2 pathways of MYB ARNT2 pathways of MYB ARNT2 MYB ARNT2 network of 90 IDCs network of 181 IDCs network in 90 IDCs network in 181 IDCs Grade 184 Grade 1954 Ribosome 45 Ribosome 55 Cell cycle 28 Cell cycle 41 LVI 178 NP 1533 VEGF 22 VEGF 35 Size 101 MC 1359 p53 17 ERBB2 29 NP 92 TF 740 ERBB2 16 p53 23 PDGFRB 13 TF 54 LYM 556 PDGFRB 22 NER 9 Proteasome 18 LNM 46 Stage 509 Proteasome 7 NER 17 MC 35 LNM 452 HR 6 MRP 10 DRS Stage 12 LVI 212 6 DRS 9 MRP 5 HR 7 LYM Age 8 161 HSA 4 BER 7 Age 1 Size 48 BER 4 HSA 6

Number of probes Number of probes Number of probes Number of probes (c) Clinical relevance (d) Biochemical profiling

Figure 3: The clinical and/or cohort significance of inferred transcriptional activities of MYB and ARNT2.ItisfeaturedbyMYB clusters (a), MYB ARNT2 transcriptional regulatory networks (b), overlapped gene pools with clinical relevance (c), and overlapped gene pools with the biochemical profiling (d). The signal transduction pathways are derived from Kyoto Encyclopedia of Genes and Genomes (KEGG) database and National Center for Biotechnology Information (NCBI) pathway interaction database. Computational and Mathematical Methods in Medicine 9

The mRNA levels of The mRNA levels of TF1 and TF2 in a other genes in a selected population selected population Population 1 Population 2 Population 3

Kaplan-Meier analysis

Univariate overlapping network Bivariate network of of TF1 and TF2 TF1nTF2 Population 1 (pCIDUGPCC TF1≤ 0.05 and ≤ pCIDUGPCC TF2 ≤ 0.05 (pCID TF1nTF2 0.05) Type 3 Type 1

M2 U Type 4 M1 Population 2 Population 3 M3 M4 Type 2

U + M1 + Type i M2 + M3 (i=1,2, 3, 4) + M4

Feature type k (k=I, II, III, IV)

Figure 4: Classification on prognostically relevant subpools of genes in a TF1 TF2 network of the selected population. Step A. We established a full network of TF1 and TF2 consisting of both univariate and multivariate portions of the network in a population (cohort 3). TF2 is an obligate transcription factor partner of TF1 in this case. Therefore, when the gene pools are gathered for further analysis (Step B), they include the gene pools from U and M1–M4 that are demonstrated by a Venn diagram. M1–M4 stand for the expressions of four gene pools following four inferred regulatory mechanisms due to the combinatorial interactions between TF1 and TF2, respectively [11]. U stands for the gene pool in the overlapping network of TF1 and TF2 but without the gene pool derived from M1. Step B. We performed the genome-wide analysis of human breast tumor gene profiles in three cohorts of interest for their prognostic relevance using Kaplan-Meier survival analysis. The prognostic relevant gene pools in three cohorts have four types based on the differential relevance among cohorts 1, 2 and 3foreach prognostic predictor of interest. Each prognostic factor can be grouped into one of four types. We define them as Type i (𝑖= 1,2,3,and4). “S” means significant. “NS” means insignificant. Types 1, 2, 3, and 4 are (S, S, S), (NS, S, S), (S, NS, S), and (NS, NS, S), respectively. A Venn diagram demonstrates the subpools of genes in relation with their prognostic relevance in at least one of three selected cohorts. Finally, the overlapped gene subpool between prognostic predictors of a given type and the full network of TF1 TF2 identified in cohort 3 is illustrated by a Venn diagram. In this study, the prognostic relevant genes in a full network of TF1 TF2 can be classified into four feature types. They are designated as feature type 𝑘 (𝑘=I, II, III, and IV). http://dx.doi.org/10.1155/2014/813067;Figures3(c) and 3(d)). MYB are demonstrated by down-regulating a set of genes Histological grade is predicted to be heavily regulated by that are predicted to attenuate the pathophenotypic devel- MYB while comparing to nine other clinical parameters in opment of breast tumors (Figures S6.7–S6.9). For instance, both 90A cohort and 181A cohort (Figures 1(a), 1(b),and the progression of histological grade, LVI, and tumor size 3(c)). In addition, the cancer-related signal transduction are suppressed in a subset of MYB highly expressed breast pathway (STP) for ribosome has more genes to be the inferred tumors showing relatively low histological grade, LVI and components of MYB network than other twelve STPs do tumor size. This may be due to the actions of some inferred (Figure 3(d)). gene components in the transcriptional regulatory network of Weobserved that increased MYB expression may be asso- MYB ARNT2.Importantly,ARNT2, MYB, XBP1, and SALL2 ciated with relatively early disease development, non-tumor are the candidate drivers for attenuating histological grade component, and 41 prognostic relevant MYB inferred target promotion (Figure S6.7). XBP1, SALL2, POU2F1, ARNT2, probes (Figures S8.1-S8.2 and Figures 1(a) and 1(b)). MYB and MYB are the candidate drivers in preventing LVI pro- is predicted to suppress the expressions of key components gression (Figure S6.8). MYB, ARNT2, and POU2F1 are the in cancer-related signal transduction pathways, such as cell potential drivers in attenuating the tumor size progression cycle, p53, PDGFRB, ERBB2 and VEGF (Figures S6.1–S6.3 (Figure S6.9). Based on the brief analysis on the tumor and S6.5-S6.6). In addition, tumor suppressive activities of suppressive activities of MYB described above, it remains 10 Computational and Mathematical Methods in Medicine

Pie distribution (%) for four gene subpools of 131 Pie distribution (%) for four gene subpools of 302 prognostic predictors prognostic predictors

1% 0%

37% 48% 46%

62%

1%

5% I III I III II IV II IV (a) (b)

Figure 5: The pie chart for feature distribution of prognostic relevant genes in the network of MYB ARNT2. First, two networks of MYB ARNT2 with cohort relevance (90A cohort and 181A cohort) identify 131 probes and 302 probes to be the potential prognostic factors in 90A cohort and 181A cohort, respectively. Second, the pie distribution made for four classified prognostic predictor subpools derived from overlapping between the cohort network of MYB ARNT2 and four types of prognostic indicators in three selected populations (91A cohort, 90A cohort, and 181A cohort), respectively. The pie chart demonstrates the sizes of four gene subpools for the given gene pool byits corresponding percentage to be distributed in a pie. The common trend shared by two pie charts is that size of gene subpools in four feature types followed an order of type IV > type II > type III > type I. largelyunknownwhetherornotMYB suppresses the gene expression is up-regulated in tumor component, which is expressions of the risk factors, which are responsible for poor consistent with the report [20]. Liu et al. [21]reportedthat clinical outcome. However, we observed that the favorable ARNT2 dimerizes with SIM1 to up-regulate their down- prognostic feature of MYB may be due to partially down- stream target genes (268 probes) in vitro,whicharepredicted regulating gene expressions for epithelial-to-mesenchymal to be functional in seven categories—transcription regula- transition (EMT) markers predicted by network analysis tors, signaling components, metabolic enzymes, channels and (Figure S6.10). Moreover, the EMT activities were predicted transporters, cell adhesion and migration, miscellaneous and to be partially suppressed by SALL2. SALL2 is a gene com- uncharacterized. Partial results of the supervised network ponent of 41-gene signature and is a putative shared target analysis in Tables S1.3 and S1.4 show MYB and ARNT2 shared gene of MYB and ARNT2. EMT related genes are involved target genes (57 genes) overlapping with the downstream in the program of development of cancer cells characterized target genes of ARNT2/SIM1 (Table S6.1). The MYB network by loss of cell adhesion, repression of E-cadherin expression, predicts ARNT2 to be a target gene of MYB in 90A cohort and increased cell mobility for promoting tumor metastasis. (Figure S6.10). ARNT2 and MYB share the large pools of targetgenes(7,225probesinTableS1.3and5,308probesin Table S1.4). A relatively lower amount of genes is putative 3.1.2. The Potential Clinical Impact of ARNT2 as the Obligate target genes of ARNT2, which dimerizes with its essential Transcription Factor Partner of MYB and Its Relation with TF partners. They are predicted to be not co-regulated by Favorable Prognostic Features of MYB. Aryl-hydrocarbon MYB (e.g., 2,322 probes of 152 ARNT2 in Table S1.3 and receptor nuclear translocator 2 (ARNT2) was identified as 3,962 probes of 120 ARNT2 in Table S1.4). This suggests the a homologue with a high degree of sequence similarity prognostic relevance of ARNT2 alone to be less significant to Aryl-hydrocarbon receptor nuclear translocator (ARNT) than that of MYB and ARNT2 in our model system. [17]. ARNT2 is a transcription factor. Human ARNT2 cDNA MYB and ARNT2 maymutuallyinteractwitheachother was identified by Barrow et al.18 [ ]. The actual function of in regulating their shared target genes during early tumor ARNT2 in cancer still remains largely unknown. Qin et al. development and co-contribute to favorable prognosis indi- [19]reportedARNT2 affecting HIF1 regulatory signaling catedinboth90Acohortand181Acohort(FiguresS8.1and and in human breast cancer cell model. It is a S8.2). The supporting pieces of evidences are as follows. First, potential favorable prognostic factor in breast cancer when we observed MYB and one of ARNT2 probes sharing the it is elevated and it is expressed higher in tumor component clinical impact on early development of histological grade, than in non-tumor component [20]. Our finding shows that mitotic count, nuclear pleomorphism and tubule formation ARNT2 is not a prognostic indicator in both 90A cohort and in 181A cohort (Figures 1(b) and 2(d)). Second, both MYB and 181A cohort (III and IV of Figure S5.2). However, its mRNA ARNT2 are determinants for the early development of LVI, G, Computational and Mathematical Methods in Medicine 11

NT AB CATSPER2 APOM SALL2 ZNF228 STK36 IQCK ANAPC4 NT C D CR621710 CATSPER2 TTC19 APOM PMS2CL SALL2 ENST00000370040 ZNF228 POU2F1 STK36 POU2F1 IQCK XBP1 ANAPC4 BC009926 CR621710 PAH TTC19 PPP1R9A PMS2CL ABAT ENST00000370040 TBC1D9 POU2F1 THSD4 POU2F1 NTN4 XBP1 TMC5 BC009926 MAP1S PAH GNAI2 PPP1R9A CCDC124 ABAT PICK1 TBC1D9 SLC25A1 THSD4 RAB42 NTN4 PKMYT1 TMC5 ACOT7 MAP1S DUSP7 GNAI2 PCBP3 CCDC124 NFKBIL2 PICK1 TOMM40 SLC25A1 ZNF598 RAB42 EIF5A PKMYT1 GAPDH ACOT7 TGM2 DUSP7 MAF1 PCBP3 GNB2 NFKBIL2 RANGAP1 TOMM40 ZNF598 E2F1 EIF5A MYBL2 GAPDH MYBL1 TGM2 ARNT2 MAF1 ARNT2 GNB2 MYB RANGAP1 E2F1 5423 5425 5293 1593 1539 1541 1543 5407 5403 5429 5999 5411 5421 5405 5419 5413 5415 5431 4425 4423 1435 1449 1531 1261 1513 1353 1331 1707 2431 2445 2427 2415 2435 5303 5309 5331 1653 5299 5295 1597 1647 MYBL2 MYBL1 ARNT2 ARNT2 MYB −0.7 −0.41−0.24 −0.11 0 0.12 0.26 0.45 0.79 5435 5423 5425 5293 1591 1593 1539 1541 1543 5431 5415 5417 5433 5407 5403 5429 5999 5409 5427 5405 5437 5419 5413 5411 5421 4425 4423 1435 1449 1261 1531 2431 1331 1707 2445 2427 2435 2415 1513 1353 1431 5323 4397 4387 5313 4369 5345 5343 5299 5305 5303 5325 5309 5329 1579 1665 1597 1619 4389 1621 1647 5347 5321 4401 4381 4433 1661 1711 4391 4379 5335 1509 5295 5341 5331 1653 (a) (b) 1 P < 0.001 1 P = 0.017

0.8 0.8

0.6 0.6

0.4 0.4

Surviving fraction 0.2 Surviving fraction 0.2

0 0 0 5 10 15 20 0 5 10 15 20 Time to death (years) Time to death (years) Subcohort A(15A) Subcohort C(16A) Subcohort B(8A) Subcohort D(35A) (c) (d)

Figure 6: In vivo validation on a favorable prognostic signature in subsets of ER(+) IDCs and 181 IDCs. Panel A presents the heatmap displaying forty-one prognostic relevant probes, which are predicted to show the consensus expression dynamics in both 90A cohort and 181A cohort. These 41 probes are part of the most relevant transcriptional activities of both MYB and ARNT2 in a subset of ER(+) IDCs (a) or in the subset of 181 IDCs predominantly containing ER(+) subtype (b). Panel B demonstrates that the Kaplan-Meier survival curves significantly show the difference in clinical outcomes when comparing subcohorts A(15A)/B(8A) (c) and subcohorts C(16A)/D(35A) (d), respectively. It indicates forty-one probes to be the favorable prognostic signature driven by MYB in coupling with ARNT2. The tumor samples in subcohorts A/C have relatively higher mRNA levels of MYB and ARNT2 than those in subcohorts B/D. However, the tumor samples in subcohorts A/C have relatively lower mRNA levels of MYBL1 and MYBL2 than those in subcohorts B/D. “NT” stands for non-tumor component.

TF, NP, size and LNM in 90A cohort (Figures 1(a), 2(a) and 3.2. The Classification of Prognostic Relevant MYB ARNT2 2(b)). Third, the networks of MYB ARNT2 predict relatively Subnetworks via Their Relevance in Predicting the Clinical low activities of the cancer-related signaling pathways due Outcome in the Specific Cohort(s). MYB is a predictor of to not regulating key oncogenic signaling molecules or favorable prognosis in 181A cohort (Figure 1(e)). It is likely suppressing the oncogenic signaling molecules (Figures S6.1– that the overall clinical impact of MYB in this cohort may S6.6). Fourth, the clinically relevant and cohort enriched serve as a determining factor for the good clinical outcome. networks of MYB ARNT2 may participate in breast cancer Each tumor sample has the unique network of development only at early phase. For instance, relatively MYB ARNT2. Based on the methodology used to establish high levels of both MYB and ARNT2 in the breast tumor the inferred network of MYB ARNT2,themostrelevant components show their clinicopathological features with low tumor suppressive activities of this network are determined grade; LVI negative and LYM negative (Figures S6.7–S6.9) in mainly from the 1/10th tumor sample group that has a subset of patients in 90A cohort. relatively high mRNA levels of both MYB and ARNT2 in a 12 Computational and Mathematical Methods in Medicine givencohorttocontributethehighestsubCIDvaluebasedon Table 1: Classification of four prognostic relevant gene subpools CID subgrouping strategy [11]. We analyzed the components within the network of MYB ARNT2. Two sets of results are for 90A of this network genome-wide to further evaluate if these cohort and 181A cohort, respectively. network components serve as prognostic indicators when each of them is expressed at a level within top ten percent (a) 90A cohort (i.e., elevated) in a population of interest via Kaplan-Meier Featuretype 90Acohort 91Acohort 181Acohort survival analysis [15]. Based on this screening strategy, I000 only limited amounts of probes are found to be significant II48048 (𝑃 ≦ 0.05)(Table 1 and Figure 5). The significance for each probe of interest in predicting clinical outcome is identified III 1 1 0 either in one population or in several populations. Three IV 82 0 0 populations (90A cohort, 91A cohort and 181A cohort) were Total # probes 131 1 48 used to classify a gene pool potentially to be the prognostic (b) 181A cohort indicators in at least one of three populations. Four subpools of genes have been derived from this classification strategy Featuretype 90Acohort 91Acohort 181Acohort andhavebeendesignatedasfourfeaturetypes(Figure 4). I 222 II 140 0 140 3.3. The Most Relevant Feature for Transcriptional Regula- III 0 14 14 tory Event of MYB in Regulating Genes Responsible for the IV 0 0 146 Favorable Clinical Outcome Is across Subtypes but Enriched Total # probes 142 16 302 in ER(+) IDCs (i.e., Feature Type II). MYB mRNA is highly expressed in ER(+) breast cancers as compared to ER(−)ones (Figure 1(c)). It is elevated especially in group IE subtype (Figure 1(d)). Additionally, MYB is an estrogen responsive gene [22]andARNT2 is a xenoestrogen responsive gene [23]. MYB ARNT2 (Figures S8.1 and S8.2). Four selected subsets The in vitro data using the breast cancer cell model MCF-7 of patients (subcohorts A, B, C, and D) differentially express- [10, 24]supportournetworkpredictionthatARNT2 is a MYB ing these 41 probes in a consensus manner within tumor target gene. Thus, estrogen action on up-regulating activities tissues (Table S4.9 and Figure 6) were identified. We further of MYB in coupling with ARNT2 may be the major feature of validatedtheutilityof41probesinprognosisin vivo (𝑃< favorable prognosis for MYB. To address this specific event, 0.001 for subcohort A versus subcohort B; 𝑃 = 0.017 for we suspect the common gene pool shared by two networks of subcohortCversussubcohortDinFigure 6). We found a MYB ARNT2 with cohort relevance (see 90A cohort and 181 trend of increasing expression levels of MYBL1 and L2 when cohort in Figure 3(b)) may uniquely represent the prognostic MYB expression level becomes low in ER(+) subgroup and/or − feature of MYB that is not only conserved across subtypes ER( ) subgroup that results in a poor survival outcome as but also enriched in ER(+) IDCs. Only feature type II closely compared to high MYB expressing subgroup (Figures 6(a) presents this event while comparison was made among four and 6(b)). It is likely that the bottom 10% of MYB expressing feature types described below. tumors may have the transcription activities shown in subco- hortsBandD.WefoundarrayIDs5309,5343,5325,4401,1711, Total 131 probes (131/480) in the clinically significant and 4391,and5335arewithinthebottom10%ofMYB expressing 90A cohort relevant network of MYB ARNT2 are identified tumors. to be prognostic predictors in at least one of three tested cohorts (90A cohort, 91A cohort, and 181A cohort). They are 3.4. The Annotated Functions of Prognostic Relevant Genes divided into four feature types (Table S4.1–S4.4). The pie dis- in the MYB Transcriptional Regulatory Subnetwork and the tribution for these four feature types show the predominant Novel Findings. The functional annotations of these 41 probes groups falling in two cohorts—90A cohort and 181A&90A show that genes are involved in stress, ion channel, phos- cohort (Figure 5 and Table 1). phorylation, dephosphorylation, transcription, translation, G On the other hand, 302 probes (302/2,727) in the clinically signaling, and metabolism for amino acids and fatty significant and 181A cohort relevant network of MYB ARNT2 acids according to Gene References into Function (Gene RIFs areidentifiedtobeprognosticpredictorsinatleastone of NCBI), Gene Spring GX7.3.1, and the related literature. of three tested cohorts (90A cohort, 91A cohort and 181A Network analysis indicates the biochemical profiling of those cohort). They are divided into four feature types (Tables S4.5– activities (Table S3.15). The function of each probe may S4.8 of Additional file 1). The pie distribution for these four not be limited by its current annotated function. MYB may feature types shows the predominant groups falling in two differentially regulate those known physiological activities. cohorts—181A cohort and 181A&90A cohort (Figure 5 and Some transcription factors may act as the co-regulators Table 1). of MYB to regulate those cellular activities. Importantly, We further examined the heatmaps for a subpool of the clinical tumor samples were collected at a time point probes (41 probes) that is the shared subpool of probes when they were surgically removed from patients. Therefore, infeaturetypeIIofthetwocohortrelevantnetworksof further studies in model systems using time course strategy Computational and Mathematical Methods in Medicine 13

Table 2: Functional annotation and transcriptional regulation patterns of the prognostic relevant signature.

Feature Regulated by Regulated by Biological function(s) and/or cancer-related Gene symbol no. MYB MYBL1 & L2 activities 2654 APOM Up — Protein 714 IQCK Up Down SRC-3 binding protein 9472 POU2F1 Up Down TF 20014 PPP1R9A Up — Phosphatase 11991 POU2F1 Up Down TF 7164 ZFP112(ZNF228) Up Down Zinc finger protein 4051 TMC5 Up Down Transmembrane channel-like protein 9673 STK36 Up Down Serine/threonine kinase 1509 TMEM87B(CR621710) Up — Transmembrane protein 6481 TTC19 Up Down Roles in protein-protein interactions 10396 BC009926 Up Down The inner mitochondrial membrane protein 20295 PAH Up — metabolism 3096 SALL2 Up — TF and putative tumor suppressor 17750 TBC1D9 Up Down Multidrug resistance gene 1(MDR1) 20917 NTN4 Up Down Good prognostic factor 8570 CATSPER2 Up — Ion channel 6691 ANAPC4 Up Down replication 10334 THSD4 Up Down A disintegrin and metalloproteinase 9326 ABAT Up Down Aminotransferase 15762 NBPF4(ENST00000370040) Up — Undefined function 10024 XBP1 Up Down TF 10998 PMS2CL Up — Mismatch repair gene 11119 TGM2 Down — Metabolism 3069 ACOT7 Down Up 1039 GNAI2 Down — G protein 19802 PKMYT1 Down Up Kinase Candidate target for cancer treatment; 20037 GAPDH Down Up proliferation and metastasis; . 5918 MAP1S Down — Morphology; microtubule associated protein 956 CCDC124 Down Up Undefined function 21198 GNB2 Down — G protein 13619 TOMM40 Down — Enzyme A negative regulator of NFKB mediated 18840 NFKBIL2 Down — transcription; the maintenance of genome stability; may cause chemoresistance Putative Ras-related protein related to cell 4681 RAB42 Down Up proliferation 3891 MAF1 Down Up Control transcription initiation 21760 ZNF598 Down Up Undefined function 17698 EIF5A Down — Translation 7939 SLC25A1 Down Up Cellular component 7824 PCBP3 Down — Posttranscriptional activities 16475 DUSP7 Down Up Phosphatase Signaling molecule; poor prognosis; promote 17776 PICK1 Down — tumor growth G protein signaling; a new target for cancer 3073 RANGAP1 Down Up chemotherapy 14 Computational and Mathematical Methods in Medicine will be appropriate to validate the roles of MYB based on its prognostic relevant signature of MYB and ARNT2 (i.e., 41- transcriptional dynamic in relation to the predicted activities gene signature). The network analysis is a qualitative method. described above. Additionally, they are potential factors to We observed that the expression levels of MYB inferred increase patient survival rate after receiving conventional targetgenesvaryalot.Assuch,someprobesinthe41- cancer treatments and some of them are tumor suppressors gene signature are not clinically significant. For example, (Table 2). Interestingly, the annotated functions of these CR621710, TBC1D9, ZNF598 and GAPDH within the 41-gene genes are largely consistent with the published data for the signature are not clinically significant in 90A cohort (Table major functional protein groups in MYB regulated genes S5.1). PPP1R9A and GNAI2 (Table S5.2) show no significant from the human erythroleukemic cell line K562 model [25]. clinical impact in 181A cohort. Additionally, most of them The most interesting finding is the inferred target genes of (39/41) have their clinical significance to be shifted away both MYB and ARNT2 including POU2F1, SALL2, and XBP1. from the clinical characteristics of MYB and ARNT2 in They are transcription factors that are also predicted to be the 181A cohort (39/41) and in 90A cohort (37/41) (Tables S5.1- favorable prognostic predictors in both 90A cohort and 181A S5.2). The clinicopathological characteristics of subcohorts cohort (Figures S5.1 and S5.2). They are clinically relevant A, B, C, and D are partially overlapped (Table S5.3). This in early tumor development (Figure S7.1–S7.4 of Additional indicates the favorable prognostic activities of MYB and file 1). ARNT2, POU2F1, and XBP1 are in MYB signature ARNT2 to be preferentially at early tumor development but of MCF-7 [10]. XBP1 is estrogen responsive [26]. However, may be extended to the later event. Likewise, the late tumor high level of XBP1s (a splicing variant of XBP1)isassociated development overlapping with a few early clinicopathological with increased tumor growth, resistance to anti-estrogen events is found in tumor samples with both suppressed therapy, and poor patient survival [27]. SALL2 is a putative activities of MYB and ARNT2. Importantly, the annotated tumor suppressor [28]. POU2F1 is the transcription factor activities of the 41-gene signature are similar to the common for proliferation and may promote genomic instability and gene activities of MYB in vitro [25]. The transcriptional tumorigenesis in breast cancers [29]. The detailed functions dynamic of this prognostic signature has shown to be across of these TFs in breast tumor development are limited. For molecular subtypes but enriched in ER(+) IDCs (Figure 6). example, the mechanisms of how they cooperatively con- However, Table S5.4 shows those univariate COXPH analyses tribute to favorable prognosis will be the important research of subcohort A/B, subcohort A/nonA, and nine major tradi- topics for better understanding of the prognostic features of tional prognostic factors in 90A cohort to be not significant. MYB. Likewise, those of subcohort C/D, subcohort C/nonC, and The favorable prognostic feature of MYB could be simply nine major traditional prognostic factors in 181A cohort are due to these tumours being more effectively treated, for not significant. This indicates the 41-gene signature tobe instance, with Tamoxifen, or that they do not as easily under- not prognostic relevance in a subset of ER(+) IDCs showing go an EMT and metastasis. The lack of clinical treatment data transcriptional dynamic of this gene signature (i.e., subcohort inourmodeltosupportournetworkpredictionisadrawback A/B or subcohort C/D) and in those showing early tumor of this study. However, 41-gene signature has been validated development with the 41-gene signature versus other gene by others [30–33]. expression patterns of the 41-gene signature in both 90A In this study, we only found NFKB1L2 to be chemore- cohort and 181A cohort. Moreover, the 41-gene signature sistant gene [30] and it is predicted to be down-regulated is not an independent prognostic factor in subcohort A/B, by MYB. NTN4, which predicts good prognosis [31], is up- subcohort C/D, subcohort A/nonA and subcohort C/nonC regulated by MYB. PICK1, which predicts poor prognosis basedonmultivariateCOXPHanalysisinboth90Acohort [32], is down-regulated by MYB.Onthecontrary,TBC1D9, which is also known as multidrug resistance gene 1(MDR1) and 181A cohort. Importantly, only 181A cohort shows the tra- [33], is up-regulated by MYB.Itisstillearlytoconcludethe ditional prognostic factors, LVI, size, LNM, stage, and LYM, treatment option and response to the treatment based on to be prognostic relevant. Typically, LNM is the independent the 41-prognostic gene signature in vivo. First, the cohort prognostic factor when comparison was made among tested study of ours is only a training set. We need an appropriate prognostic factors. testing set to validate its reproducibility. Second, the cancer We suspect that both univariate and multivariate COX treatment data for the cohort of ours is incomplete based on proportional hazard (COXPH) analyses for this signature the medical record. Third, the in vitro and in vivo studies shownotsignificant(TableS5.4)duetotheuniqueregulatory of the gene signature at protein level and its relation to mechanisms of MYB in coupling with ARNT2 and the small 𝑁 cancer treatments would be necessary to conclude genes for number for those tested cohorts. However, based on the the cancer treatment option(s) and response(s) to cancer rationale of supervised network analysis, we observed that treatment(s). Kaplan-Meier survival analysis predicts the prognostic sig- Here, we claim that 41-gene signature is different from nificance of 41-gene signature in a subset of IDCsFigure ( 6). other published signatures due to a supervised network Furtherinvestigationsinalargepopulationtoevaluatethe analysis approach. First, each functional transcription factor reproducibility of this favorable prognostic signature would (e.g., MYB)hasitsowntranscriptionalmechanismspredicted be necessary. by network analysis. Network analysis allows dissecting MYB activities by its transcriptional regulatory network. Second, 3.5.TheClinicalRolesofMYBFamilyMembersin90ACohort a supervised network analysis has identified a potential and 181A Cohort. MYB family members—MYB, MYBL1 and Computational and Mathematical Methods in Medicine 15

ER𝛼 E2F1

MYB ARNT2 MYBL1 MYBL2

Poor Favorable Poor Favorable prognosis prognosis prognosis prognosis

RAB42 APOM PAH TGM2 RANGAP1 RAB42 IQCK TBC1D9 ACOT7 MAF1 IQCK SALL2 ACOT7 MAF1 POU2F1 NTN4 GNAI2 ZNF598 POU2F1 TBC1D9 PKMYT1 EIF5A PKMYT1 ZNF598 ∗ THSD4 PPP1R9A NTN4 POU2F1 GAPDH SLC25A1 ∗ GAPDH SLC25A1 ABAT POU2F1 CATSPER2 ZFP112 MAP1S PCBP3 CCDC124 DUSP7 TMC5 XBP1 CCDC124 DUSP7 ZFP112 ANAPC4 GNB2 PICK1 TMC5 THSD4 STK36 BC009926 RANGAP1 TOMM40 STK36 ABAT TTC19 ANAPC4 NFKBIL2 TMEM87B NBPF4 TTC19 XBP1 BC009926 PMS2CL

Figure 7: The inferred prognostic predictors partially antagonized by MYBL1 and MYBL2 during late tumor progression. This event is proposed to be differentially controlled by𝛼 ER and/or ER𝛼 E2F1 promoter use pathways. Here, the novel predicted role of MYBL1 and MYBL2 may provide part of mechanism for MYBL1 and MYBL2 being the unfavorable prognostic predictors in a subset of ER(+) IDCs. The increased levels of both MYBL1 and MYBL2 are frequently accompanied with low expression levels of MYB and ARNT2 observed in these tumor samples. There are twenty-four probes in this signature predicted to be shared target genes of both MYBL1 and MYBL2.

MYBL2 have been studied in breast cancers. But, the genome- in 90A cohort predict poor clinical outcome, respectively wide regulatory mechanisms for these TFs to their shared (Figure S5.1). The preferential poor prognostic activities of target genes in breast cancers are largely unknown. MYBL1 and MYBL2 arealsoindicatedinER(−)IDCs(𝑛= 25) of subcohort D. Typically, we found it in two ERBB2 3.5.1. The Clinical Impacts of MYB Family Members. ANOVA IDCs (array IDs 5305 and 1711) (Figure 6(b)). However, the tests on MYB for its clinical impact in 90A cohort and 181A prognostic feature of MYBL1 and L2 is less significant in − cohort suggest its role in early tumor development. However, ER( )cohort(datanotshown).Thismaybebecausethe − the expression levels of MYBL1 and MYBL2 are increased majority of those ER( ) IDCs were analyzed for their survival during later development of breast cancers. outcomes when they were less than 5 years from the first High mRNA levels of MYBL1 are the determinants of diagnosis. Therefore, the follow-up survival analysis to find − LNM,size,HER,G,NP,andMCinlatebreasttumor out the prognostic features of MYBL1 and L2 in ER( )IDCs development (Figure S7.5).Increased mRNA levels of MYBL2 will be needed in the future. are significantly associated with late LVI, PR, LNM, G, TF, The recent research evidence supports our finding that NP and MC (Figure S7.6). ANOVA test shows MYBL1 to MYB is a potential favorable prognostic factor in luminal be a promoter for the late clinicopathological progression of breast cancer [10]. Thorner et al. [6] demonstrated that ER(+), ER(−), and 181 IDCs. MYBL2 is a promoter for the late increased MYBL2 expression is a significant predictor of poor clinicopathological progression of both ER(+) IDCs and 181 survivalandpathologicalcompleteresponsetoneoadjuvant IDCs. chemotherapy (e.g., DNA topoisomerase II 𝛼 (TOP2A) inhibitors—doxorubicin and etoposide) in basal-like breast 3.5.2. The Prognostic Values of MYB Family Members. MYB cancer. The MYBL2 hasbeendiscoveredasoneofrecurrence is predicted to be a favorable prognostic indicator in 181A risk genes in tamoxifen-treated, node-negative breast cancer cohort (Figure 1(e)). However, elevated MYBL1 and MYBL2 [34]. MYBL1 is a transcription factor that is involved in 16 Computational and Mathematical Methods in Medicine mammary gland development [5]. It may play roles in the 3.6. The Undiscovered Transcriptional Activities and Interac- biology and/or pathogenesis of some neoplasia [35, 36]. tion among MYB Family Members. Our preliminary data Recent report mentioned MYBL1 to be an oncogene [37]. on the clinical roles of MYBL1 and MYBL2 in ER(−)and ER(+) breast cancers (Figures S7.5 and S7.6) suggest that 3.5.3. The Predicted Overlapping Networks of MYB Family they are important to be further investigated in the future Members in Breast Cancers in relation to Their Prognostic studies. Their interactions with MYB in different subtypes Features. Thec-Myb(MYB)proteinwasfoundtobeasso- and in different clinicopathological statuses may alter the ciated with over 10,000 promoters as the evidence of being a prognostic features of MYBL1 and MYBL2 in a subset of master transcription factor in MCF-7 cell model [24]. MYB, breast cancer population. Multiple drug targets for genes MYBL1 and MYBL2 have different regulatory mechanisms resistant to standard cancer therapies may be uncovered to but share the conserved DNA binding domain that strongly aid with the prognostication of a subset of breast cancer suggests the compensatory effects within family members in patients and with alternate treatment options at the time of regulating their shared target genes [38, 39]. As such, we diagnosis. furtheranalyzethesharedtargetgenesofMYBL1, MYBL2, and MYB in relation to their prognostic features (Table S4.10). As a result, this is the first time it was reported that MYBL1 4. Conclusions and MYBL2 may partially antagonize the action of 24 probes in the favorable prognosis signature that are predicted to MYB predicts a favorable prognosis across molecular sub- types of infiltrating ductal breast carcinomas but enriched be regulated by MYB and ARNT2 (Table 2 and Figure 7). in ER(+) IDCs. This specific event can be linked with Moreover, our data suggests that both MYBL1 and MYBL2 a 41-gene prognostic signature or a core subnetwork of are predicted to be the shared target genes of E2F1 and ER𝛼. MYB ARNT2. The supervised analysis for constructing an The promoter region of MYBL2 has E2F1/3 binding site [40]. inferred transcriptional regulatory network is efficient and However, the current report showed the major regulatory inexpensive. To our best knowledge, this signature is not the element in MYBL1 promoter region to be Sp1 sites and same as other published signatures that have been described CCAT box (a NF-Y binding site) [41]. Further investigation in a recent review [46] due to different method and the in the cell model would be of interest to validate the novel supervised approach. It is predicted to fill in the gap between findingsofoursthattheE2F1 may regulate MYBL1 expres- the traditional clinical prognostic factors and other published sion. MYBL1 and MYBL2 are estrogen responsive genes [34, prognostic signatures. However, this may be true only for 37]. They (E2F1, MYBL1, and MYBL2) are poor prognostic a subset of population (approximately 10% of a cohort) factors in 90A cohort (Figure S5.1). We, therefore, map the who obtain not only the most relevant dynamic changes of transcriptional activities of MYB family members and their gene expression pattern for the selected gene set but also gene partners during breast tumor development (Figure 7). significance in the Kaplan-Meier survival analysis. Together, Based on this hypothesized mechanism, the estrogen action such experimental design may offer the opportunity for the on MYB family members at different time point of disease personalized medicine to be discovered by the supervised development may be shown by the ER𝛼 and ER𝛼 E2F1 network analysis. promoterusepathways.Thereareforty-oneprobesasthe MYB governs a large pool of target genes based on inferred target genes of MYB and ARNT2. MYB may actively network analyses. We observed that MYB has an essential suppress oncogenic activities of NFKBIL2 [30], GAPDH [42], partner gene—ARNT2—that is low in non-tumor component RAB42 [43], EIF5A [44], and PICK1 [32]. Additionally, MYB but is up-regulated in breast tumors. Both transcription may promote good prognosis via up-regulating NTN4 and factors may coordinately suppress 13 cancer-related signal transduction pathways and some clinicopathological pro- SALL2. NTN4 is a good prognostic factor [31]. SALL2 is a < putative tumor suppressor [28]. gression ( 10 clinical parameters) in breast tumors via differ- entially regulating their shared target genes. This indicates the On the contrary, there are only twenty-four probes as major clinical impact of both MYB and ARNT2 to be tumor the candidate target genes of MYBL1 and MYBL2 (Table 2). suppressive during early tumor development. Wehavebrieflyevaluatedsomeevidence(seeTable 2)that The functional annotated 41-gene prognostic signature may support the possible poor prognostic features of MYBL1 indicates the major contributors associated with the prog- and MYBL2 and may offer new strategies in treating a subset nostic features of MYB including up-regulating the tran- of advanced ER(+) breast cancer expressing high levels of scriptional activities of three transcription factors—POU2F1, MYBL1 and MYBL2. For instance, two suggested targets for SALL2, and XBP1 which are also favorable prognostic indica- cancer treatment, GAPDH and RANGAP1,arepredictedto tors in 90A cohort and 181A cohort. Silencing both MYB and be up-regulated by MYBL2 and MYBL1.BothGAPDH and ARNT2 in 90 IDCs reveals an increase in expression levels RAB42 (the RAS oncogene family member) may be up- of some unfavorable prognostic predictors. They include regulated by these two TFs to promote oncogenic activities. E2F1, MYBL1, and MYBL2. These transcription factors may PKMTY1 is a serine/threonine protein kinase and a cell partially antagonize the favorable activities of both MYB cycle regulatory gene that is predicted to be up-regulated by and ARNT2 to lead the poor clinical outcome of a subset MYBL2 and MYBL1 suggesting increased cell proliferating of patients. Importantly, knockdown of the transcriptional activities [45]. activities of E2F1, MYBL1, and MYBL2 may be considered Computational and Mathematical Methods in Medicine 17 as the suggested treatment targets to improve prognosis for HR signaling: Signal transduction pathway of asubsetofbreastcancerpopulationwithER(−)orwith homologous recombination (HR) advanced ER(+) breast cancers who have elevated E2F1, HSΑ signaling: Signal transduction pathway of MYBL1 and MYBL2 in their breast tumors. $hsa03450$ (HSΑ) From this limited study, we only predict the major IQCK: IQ motif containing K prognostic features of MYB with in vivo validation of 41-gene LNM: Number of lymph node metastasis prognostic signature in a breast cancer model system. The LVI: Lymphovascular invasion detailed mechanisms of actions for MYB family in cancer LYM: Lymph node metastasis status development involving other transcription factor partners, MAF1: MAF1 homolog (S. cerevisiae) such as SALL2, XBP1, and POU2F1,arestillnotclear.Further MAP1S: Microtubule-associated protein 1S research to elucidate the roles of MYB family members in MC: Mitotic count breast cancers in depth is necessary, such as in the large MRP signaling: Signal transduction pathway of patient population studies and in studies using different in mismatch repair pathway (MRP) vivo and in vitro models. MYB/C MYB: v-myb myeloblastosis viral oncogene homolog (avian) MYBL1/A MYB: v-myb myeloblastosis viral oncogene homolog (avian)-like 1 Abbreviations MYBL2/B MYB: v-myb myeloblastosis viral oncogene ABAT: 4 Aminobutyrate aminotransferase homolog (avian)-like 2 ACOT7: Acyl-CoA thioesterase 7 NBPF4: Neuroblastoma breakpoint family, ANAPC4: Anaphase promoting complex member 4 subunit 4 NER signaling: Signal transduction pathway of APOM: Apolipoprotein M nucleotide excision repair (NER) ARNT2: Aryl-hydrocarbon receptor nuclear NFKBIL2: Nuclear factor of kappa light translocator 2 polypeptide gene enhancer in B-cells BER signaling: Signal transduction pathway of base inhibitor-like 2 excision repair (BER) NP: Nuclear pleomorphism CATSPER2: Cation channel, sperm associated 2 NTN4: Netrin 4 CCDC124: Coiled-coil domain containing 124 p53 signaling: Signal transduction pathway of p53 Cell cycle signaling: Signal transduction pathway of cell PAH: Phenylalanine hydroxylase cycle PCBP3: Poly(rC) binding protein 3 Cohort: A group of individuals in a PDGFRB signaling: Signal transduction pathway of population all with the features platelet-derived growth factor suitable for the study of interest or a receptor, beta polypeptide (PDGFRB) given population PICK1: Protein interacting with PRKCA 1 DR signaling: Signal transduction pathway of DNA PKMYT1: Protein kinase, membrane associated replication (DR) tyrosine/threonine 1 DUSP7: Dual specificity phosphatase 7 PMS2CL: PMS2 C-terminal like pseudogene E2F1: E2F transcription factor 1 POU2F1: POU class 2 homeobox 1 EIF5A: Eukaryotic translation initiation PPP1R9A: Protein phosphatase 1, regulatory factor 5A (inhibitor) subunit 9A ER: Estrogen receptor 𝛼 PR: Progesterone receptor ERBB2 signaling: Signal transduction pathway of Proteasome: Signal transduction pathway of v-erb-b2 erythroblastic leukemia proteasome viral oncogene homolog 2, RAB42: RAB42, member RAS oncogene neuro/glioblastoma derived family oncogene homolog (avian) (ERBB2) RANGAP1: RanGTPaseactivatingprotein1 ESR1: Estrogen receptor 1 Ribosome: Signal transduction pathway of GAPDH: Glyceraldehyde-3-phosphate ribosome dehydrogenase SALL2: Sal-like 2 (Drosophila) GNAI2: Guanine nucleotide binding protein Size: Tumor size (G protein), alpha inhibiting activity SLC25A1: Solute carrier family 25 polypeptide 2 (mitochondrial carrier; citrate GNB2: Guanine nucleotide binding protein transporter), member 1 (G protein), beta polypeptide 2 STK36: Serine/threonine kinase 36, fused Grade: Histological grade homolog (Drosophila) Group IE: ER(+)PR(+) TBC1D9: TBC1 domain family, member 9 Group IIE: ER(+)PR(−) (with GRAM domain) HER: HER-2/neu TF: Tubule formation 18 Computational and Mathematical Methods in Medicine

TGM2: Transglutaminase 2 (C polypeptide, [5] A. Toscani, R. V. Mettus, R. Coupland et al., “Arrest of sper- protein-glutamine-gamma- matogenesis and defective breast development in mice lacking glutamyltransferase) A-myb,” Nature,vol.386,no.6626,pp.713–717,1997. THSD4: Thrombospondin, type I, domain [6] A. R. Thorner, K. A. Hoadley, J. S. Parker, S. Winkel, R. C. containing 4 Millikan, and C. M. Perou, “In vitro and in vivo analysis of B- TMC5: Transmembrane channel-like 5 Myb in basal-like breast cancer,” Oncogene,vol.28,no.5,pp. TMEM87B: Transmembrane protein 87B 742–751, 2009. TOMM40: Translocase of outer mitochondrial [7] A. Sala, “B-MYB, a transcription factor implicated in regulating membrane 40 homolog (yeast) cell cycle, apoptosis and cancer,” European Journal of Cancer, TTC19: Tetratricopeptide repeat domain 19 vol. 41, no. 16, pp. 2479–2484, 2005. VEGF signaling: Signal transduction pathway of [8]E.V.Kalashnikova,A.S.Revenko,A.T.Gemoetal., “ANCCA/ATAD2 overexpression identifies breast cancer vascular endothelial growth factor patients with poor prognosis, acting to drive proliferation and (VEGF) survival of triple-negative cells through control of B-Myb and XBP1: X-box binding protein 1 EZH2,” Cancer Research,vol.70,no.22,pp.9402–9412,2010. ZFP112: Zinc finger protein 112 homolog [9] R. G. Ramsay and T. J. Gonda, “MYB function in normal and (mouse) cancer cells,” Nature Reviews Cancer,vol.8,no.7,pp.523–534, ZNF598: Zinc finger protein 598. 2008. [10] A. R. Thorner, J. S. Parker, K. A. Hoadley, and C. M. Perou, Conflict of Interests “Potential tumor suppressor role for the c-Myb oncogene in luminal breast cancer,” PLoS ONE,vol.5,no.10,ArticleID The authors declare that they have no conflict of interests. e13073, 2010. [11] L. Y. D. Liu, L. Y. Chang, W. H. Kuo et al., “In silico prediction for regulation of transcription factors on their shared target Authors’ Contribution genes indicates relevant clinical implications in a breast cancer population,” Cancer Informatics, vol. 11, pp. 113–137, 2012. Li-Yu D. Liu performed the computational analysis and [12] L. Y. D. Liu, L. Y. Chang, W. H. Kuo et al., “Major functional prepared illustrations, figures, and tables for this paper. Wen- transcriptome of an inferred center regulator of an ER(-) breast HungKuocollectedandconfirmedsurvivaldataof181 cancer model system,” Cancer Informatics, vol. 11, pp. 87–111, patients for this study. Li-Yun Chang designed the study, 2012. performed Venn diagram analyses, and drafted the paper. All [13] L. Y. Liu, C. Y. Chen, M. J. Chen et al., “Statistical identification theauthorsreadandapprovedthefinalpaper. of gene association by CID in application of constructing ER regulatory network,” BMC Bioinformatics,vol.10,article85, 2009. Acknowledgments [14] W.-H. Kuo, L.-Y. Chang, D. L.-Y. Liu et al., “The interactions between GPR30 and the major biomarkers in infiltrating ductal The financial support of this work was mainly from Grants carcinoma of the breast in an Asian population,” Taiwanese (NSC95-2314-B-002-255-MY3 and NSC98-2314-B-002-093- Journal of Obstetrics and Gynecology,vol.46,no.2,pp.135–145, MY2) (to Dr. Fon-Jou Hsieh). The authors feel thankful 2007. to receive some technical support from Welgene Biotech [15]P.S.Ray,J.Wang,Y.Quetal.,“FOXC1isapotentialprognostic Company in Taiwan. In addition, they owe many thanks to biomarker with functional significance in basal-like breast the great assistance from the office of medical record (Cancer cancer,” Cancer Research, vol. 70, no. 10, pp. 3870–3876, 2010. Registry, Medical Information Management Office, NTUH) [16]G.Zheng,K.Tu,Q.Yangetal.,“ITFP:anintegratedplatformof for accessing medical records of those patients who agreed mammalian transcription factors,” Bioinformatics,vol.24,no. on providing their specimens for microarray study. 20, pp. 2416–2417, 2008. [17] K. Hirose, M. Morita, M. Ema et al., “cDNA cloning and tissue- specific expression of a novel basic helix-loop-helix/PAS factor References (Arnt2) with close sequence similarity to the aryl hydrocarbon [1] Cancer Genome Atlas Network, “Comprehensive molecular receptor nuclear translocator (Arnt),” Molecular and Cellular portraits of human breast tumours,” Nature,vol.490,no.7418, Biology,vol.16,no.4,pp.1706–1713,1996. pp.61–70,2012. [18]L.L.Barrow,M.E.Wines,P.A.Romitti,B.C.Holdener,and J. C. Murray, “Aryl hydrocarbon receptor nuclear translocator [2]R.Y.Miao,Y.Drabsch,R.S.Crossetal.,“MYBisessentialfor 2(ARNT2):structure,genemapping,polymorphisms,and mammary tumorigenesis,” Cancer Research,vol.71,no.22,pp. candidate evaluation for human orofacial clefts,” Teratology,vol. 7029–7037, 2011. 66,no.2,pp.85–90,2002. [3] J.S.Lipsick,“TheC-MYBstory—isitdefinitive?”Proceedings of [19] X.-Y. Qin, F. Wei, J. Yoshinaga, J. Yonemoto, M. Tanokura, and the National Academy of Sciences of the United States of America, H. Sone, “SiRNA-mediated knockdown of aryl hydrocarbon vol. 107, no. 40, pp. 17067–17068, 2010. receptor nuclear translocator 2 affects hypoxia-inducible factor- [4] J. A. Rosinski and W. R. Atchley, “Molecular evolution of the 1 regulatory signaling and metabolism in human breast cancer Myb family of transcription factors: evidence for polyphyletic cells,” FEBS Letters,vol.585,no.20,pp.3310–3315,2011. origin,” Journal of Molecular Evolution,vol.46,no.1,pp.74–83, [20]V.Martinez,S.Kennedy,P.Doolanetal.,“Drugmetabolism- 1998. related genes as potential biomarkers: analysis of expression in Computational and Mathematical Methods in Medicine 19

normal and tumour breast tissue,” Breast Cancer Research and [35] J. Golay, M. Luppi, S. Songia et al., “Expression of A-myb, Treatment,vol.110,no.3,pp.521–530,2008. but not c-myb and B-myb, is restricted to Burkitt’s lymphoma, [21] C. Liu, E. Goshu, A. Wells, and C.-M. Fan, “Identification of the sIg+ B-acute lymphoblastic leukemia, and a subset of chronic downstream targets of SIM1 and ARNT2, a pair of transcription lymphocytic leukemias,” Blood,vol.87,no.5,pp.1900–1911, factors essential for neuroendocrine cell differentiation,” Jour- 1996. nalofBiologicalChemistry,vol.278,no.45,pp.44857–44867, [36] J. Golay, V. Facchinetti, G. Ying, and M. Introna, “The A-myb 2003. transcription factor in neoplastic and normal B cells,” Leukemia [22] Y. Drabsch, H. Hugo, R. Zhang et al., “Mechanism of and Lymphoma,vol.26,no.3-4,pp.271–279,1997. and requirement for estrogen-regulated MYB expression in [37] B. E. Bachmeier, V. Mirisola, F. Romeo et al., “Reference estrogen-receptor-positive breast cancer cells,” Proceedings of profile correlation reveals estrogen-like trancriptional activity the National Academy of Sciences of the United States of America, of curcumin,” Cellular Physiology and Biochemistry,vol.26,no. vol. 104, no. 34, pp. 13762–13767, 2007. 3,pp.471–482,2010. [23] X.-Y. Qin, H. Zaha, R. Nagano, J. Yoshinaga, J. Yonemoto, [38] J. J. Rushton and S. A. Ness, “The conserved DNA binding and H. Sone, “Xenoestrogens down-regulate aryl-hydrocarbon domain mediates similar regulatory interactions for A-Myb, B- receptor nuclear translocator 2 mRNA expression in human Myb, and c-Myb transcription factors,” Blood Cells, Molecules, breast cancer cells via an estrogen receptor alpha-dependent and Diseases,vol.27,no.2,pp.459–463,2001. mechanism,” Toxicology Letters,vol.206,no.2,pp.152–157,2011. [39] S. Bergholtz, T. Ø. Andersen, K. B. Andersson, J. Borrebæk, B. [24]A.M.Quintana,F.Liu,J.P.O’Rourke,andS.A.Ness, Luscher,¨ and O. S. Gabrielsen, “The highly conserved DNA- “Identification and regulation of c-Myb target genes in MCF-7 binding domains of A-, B- and c-Myb differ with respect to cells,” BMC Cancer, vol. 11, article 30, 2011. DNA-binding, phosphorylation and redox properties,” Nucleic [25] P.I. Lorenzo, E. M. Brendeford, S. Gilfillan et al., “Identification Acids Research,vol.29,no.17,pp.3546–3556,2001. of c-Myb target genes in K562 cells reveals a role for c-Myb as [40] N. Hanada, H.-W.Lo, C.-P.Day, Y. Pan, Y. Nakajima, and M.-C. amasterregulator,”Genes and Cancer,vol.2,no.8,pp.805–817, Hung, “Co-regulation of B-Myb expression by E2F1 and EGF 2011. receptor,” Molecular Carcinogenesis,vol.45,no.1,pp.10–17, [26]S.Sengupta,C.G.Sharma,andV.C.Jordan,“Estrogenregula- 2006. tion of X-box binding protein-1 and its role in estrogen induced [41] V. Facchinetti, R. Lopa, F. Spreafico et al., “Isolation and growth of breast and endometrial cancer cells,” Hormone characterization of the human A-myb promoter: regulation by Molecular Biology and Clinical Investigation,vol.2,no.2,pp. NF-Y and Sp1,” Oncogene,vol.19,no.34,pp.3931–3940,2000. 235–243, 2010. [42] C. Guo, S. Liu, and M.-Z. Sun, “Novel insight into the role of [27] M.P.A.Davies,D.L.Barraclough,C.Stewartetal.,“Expression GAPDH playing in tumor,” Clinical and Translational Oncology, and splicing of the unfolded protein response gene XBP-1 are vol.15,no.3,pp.167–172,2013. significantly associated with clinical outcome of endocrine- [43]H.E.Gruber,G.L.Hoelscher,andE.N.HanleyJr.,“Annulus treated breast cancer,” International Journal of Cancer,vol.123, cells from more degenerated human discs show modified gene no. 1, pp. 85–88, 2008. expression in 3D culture compared with expression in cells from [28] H. Liu, A. S. Adler, E. Segal, and H. Y. Chang, “Atranscriptional healthier discs,” Spine Journal,vol.10,no.8,pp.721–727,2010. program mediating entry into cellular quiescence,” PLoS Genet- [44]M.Caraglia,M.H.Park,E.C.Wolff,M.Marra,andA. ics,vol.3,no.6,articlee91,2007. Abbruzzese, “eIF5A isoforms and cancer: two brothers for two [29] C. Zhou, Y.Tong, K. Wawrowsky, S. Bannykh, I. Donangelo, and functions?” Amino Acids,vol.44,no.1,pp.103–109,2013. S. Melmed, “Oct-1 induces pituitary tumor transforming gene expression in endocrine tumors,” Endocrine-Related Cancer, [45] J. Jiang and D. Sliva, “Novel medicinal mushroom blend vol.15,no.3,pp.817–831,2008. suppresses growth and invasiveness of human breast cancer cells,” International Journal of Oncology,vol.37,no.6,pp.1529– [30] W. Piwko, M. H. Olma, M. Held et al., “RNAi-based screening 1536, 2010. identifies the Mms22L-Nfkbil2 complex as a novel regulator of DNA replication in human cells,” EMBO Journal,vol.29,no.24, [46] M. C. Abba, E. Lacunza, M. Butti, and C. M. Aldaz, “Breast pp.4210–4222,2010. cancer biomarker discovery in the functional genomic age: a systematic review of 42 gene expression signatures,” Biomarker [31] S. Esseghir, A. Kennedy, P. Seedhar et al., “Identification of Insights,vol.5,pp.103–118,2010. NTN4, TRA1, and STC2 as prognostic markers in breast cancer in a screen for signal sequence encoding ,” Clinical Cancer Research, vol. 13, no. 11, pp. 3164–3173, 2007. [32] B. Zhang, W. Cao, F. Zhang et al., “Protein interacting with C 𝛼 kinase 1 (PICK1) is involved in promoting tumor growth and correlates with poor prognosis of human breast cancer,” Cancer Science, vol. 101, no. 6, pp. 1536–1542, 2010. [33]Y.Li,P.W.Yan,X.E.Huang,andC.G.Li,“MDR1gene C3435Tpolymorphism is associated with clinical outcomes in gastric cancer patients treated with postoperative adjuvant chemotherapy,” Asian Pacific Journal of Cancer Prevention,vol. 12,no.9,pp.2405–2409. [34] S. Paik, S. Shak, G. Tang et al., “A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer,” The New England Journal of Medicine, vol. 351, no. 27, pp. 2817– 2826, 2004. Copyright of Computational & Mathematical Methods in Medicine is the property of Hindawi Publishing Corporation and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.